ADP-Ribosylation Factor 6 Regulates a Novel Plasma Membrane Recycling Pathway

HeLa cells were grown in DME supplemented with 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, and 500 μg/ml gentamycin sulfate at 37°C with 5% CO₂. Rabbit polyclonal antibodies were raised against a COOH-terminal peptide of ARF6, residues 164–175. The antiserum specifically recognizes ARF6 in transfected cells and on immunoblots; ARF6-specific labeling during immunofluorescence or immunoblotting was blocked by the immunizing peptide. A mouse mAb (16B12) against the influenza HA epitope was purchased from BabCo (Berkeley, CA). The human Tac antigen (IL-2 receptor α subunit) was detected with mouse monoclonal anti-Tac antibodies (7G7; Rubin et al., 1985). Rabbit antibodies to human transferrin were purchased from Boehringer Mannheim Biochemicals (Indianapolis, IN). Mouse antibodies against human MHC class I, W6/32 were kindly provided by Dr. Paul Roche (National Institutes of Health, Bethesda, MD). Fluorescein-conjugated WGA and Oregon green–labeled phalloidin were obtained from Molecular Probes, Inc. (Eugene, OR). Fluorescein- and rhodamine-conjugated donkey anti–mouse and donkey anti–rabbit IgG were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA). All other reagents, including iron-saturated human transferrin, CD, and latrunculin B, were purchased from Sigma Chemical Co. (St. Louis, MO).

PCR (Perkin-Elmer Cetus Instruments, Norwalk, CT) was used to remove DNA sequences encoding a COOH-terminal HA epitope tag from the cDNAs encoding human ARF6/T27N and ARF6/Q67L generated previously (Peters et al., 1995). A DNA fragment from the ClaI site at 286 bp to the end of ARF6, containing a stop codon and a BglII site, was amplified by PCR and digested with ClaI/BglII. This fragment was used to replace the corresponding fragment in the epitope-tagged T27N and Q67L mutants of ARF6. The sequences of all constructs, including regions generated by PCR, were confirmed by DNA sequencing (SeqWright DNA Sequencing, Houston, TX). A modified pCDL-SRα expression vector (termed pXS; Takebe et al., 1988) was used to express wild-type and mutant ARF6, as well as Tac cDNAs in HeLa cells.

HeLa cells were grown on glass coverslips and transfected using the calcium phosphate method as described previously (Bonifacino et al., 1989). Cotransfections of ARF6 and Tac were performed using 10 μg of each plasmid in 10-cm dishes. Under these conditions, the level of ARF6 overexpression was estimated to be 10–40-fold over the endogenous protein.

30–36 h after transfection, transfected cells were treated as indicated, fixed with 2% formaldehyde in PBS for 10 min at room temperature, and rinsed with 10% FBS and 0.02% azide in PBS (PBS/serum). Cells were incubated with primary antibodies diluted in PBS/serum containing 0.2% saponin for 1 h at room temperature and were washed (three times, 5 min each) with PBS/serum. The cells were then incubated with fluorescently labeled secondary antibodies diluted in PBS/serum plus 0.2% saponin for 1 h, washed again, and mounted on glass slides.

Transfected cells were rinsed briefly three times with 0.5% BSA in DME, and then incubated in the same medium for 30 min at 37°C in the absence or presence of the indicated drugs. Cells were then incubated with 30 μg/ ml of iron-saturated human transferrin in the continued presence of drugs for an additional 30 min at 37°C, rinsed quickly three times with DME containing 10% FBS, fixed, and processed for immunofluorescence as above.

For Tac antibody internalization, cells cotransfected with Tac and ARF6 constructs were chilled to 4°C and incubated with mouse anti-Tac antibodies for 30 min in an ice/water bath. The cells were rinsed briefly with ice-cold DME containing 10% FBS (complete medium), and they were incubated with prewarmed (37°C) complete medium in the absence or presence of the indicated drugs for 30 min at 37°C. Cells were then fixed either immediately or after washing with low pH buffer (Klausner et al., 1983). To remove anti-Tac antibody remaining at the surface, the cells were chilled to 4°C, rinsed three times quickly with 0.5% acetic acid, 0.5 M NaCl, pH 3.0 (low pH buffer), and then three times with ice-cold complete medium before fixation.

To detect ARF6 and the internalized Tac antibody, fixed cells were labeled with rabbit anti-ARF6 antiserum, washed, and then incubated with fluorescently labeled donkey anti–rabbit IgG to detect ARF6 staining, and incubated with the appropriate fluorescently labeled donkey anti– mouse IgG to detect the internalized Tac antibody. To detect Tac antibody remaining at the cell surface, some cells were incubated with fluorescently labeled donkey anti–mouse IgG in the absence of saponin; the cells were then washed and incubated in the presence of saponin with anti-ARF6 antibodies, followed by the appropriate fluorescently labeled donkey anti–rabbit IgG.

To monitor the recycling of Tac antibody, cells cotransfected with Tac and ARF6 constructs were labeled with mouse anti-Tac antibodies at 4°C as described above, and were then warmed to 37°C for 30 min either in the absence or presence of 1 μM CD. The cells were chilled to 4°C, quickly washed with low pH buffer, as described above, to remove Tac antibody remaining at the cell surface, and were warmed again to 37°C for 30 min in the absence or presence of CD before fixation. The surface reappearance of Tac antibody was detected by incubating the fixed cells with fluorescently labeled donkey anti–mouse IgG in the absence of saponin; ARF6 was then labeled with anti-ARF6 antiserum and the appropriate secondary antibodies in the presence of saponin.

We previously observed that inhibition of actin polymerization, by treatment of cells with CD, inhibited the AlF-induced protrusions in HeLa cells overexpressing ARF6, and additionally, we noticed that CD treatment shifted the distribution of ARF6 to an internal membrane compartment (Radhakrishna et al., 1996). To further investigate this response to CD, HeLa cells transfected with plasmid encoding ARF6 were treated with CD for 30 min. The cells were then fixed and immunolabeled with polyclonal anti-ARF6 antibodies, and the actin filaments were labeled with Oregon green–conjugated phalloidin. In untreated cells, ARF6 localized along the PM at the peripheral edge of cells and in internal structures near the nucleus. Fine tubular elements were observed emanating out of the juxtanuclear region towards the peripheral edge of some cells (Fig. 1, untreated). The extent to which these tubular elements were observed in untreated cells varied (for example, see tubular elements in Figs. 4 and 5); often they were difficult to discern over the cell surface labeling.

A striking enhancement of ARF6 labeling of the tubular structures was observed in cells treated with 1.0 μM CD (Fig. 1, CD). Morphologically, these tubular structures were similar to those observed in the absence of CD, but they were more extensive, suggesting that CD treatment was not creating but merely shifting more ARF6 to these structures. Other inhibitors of actin polymerization, cytochalasin B and latrunculin B (Spector et al., 1989), also resulted in extensive tubular staining for ARF6 (not shown). The effects of CD treatment on ARF6 distribution were fully reversible within 30 min after the removal of CD (not shown). Even 0.1 μM CD was sufficient to shift ARF6 to the tubular structures and inhibit the AlF response, although little change in the appearance of actin filaments was evident (data not shown).

These tubular elements were internal structures, and not invaginations of the PM, as recently described (van Deurs et al., 1996). Surface staining of fixed cells, in the absence of permeabilization, with fluorescent WGA (not shown) or antibodies to PM proteins (e.g., see Fig. 5) did not label the tubular structures. Furthermore, we could demonstrate that these were internal structures that could be loaded from the PM and thus represented an endosomal compartment (see Fig. 5). Double labeling with antibodies to ARF6 and tubulin showed that the tubular compartment was aligned along the microtubules; the ability to observe these tubular structures during CD treatment was diminished when microtubules were disrupted by pretreatment with nocodazole (data not shown).

Immunoelectron microscopy had previously shown that wild-type ARF6 localized to both the PM and internal tubulovesicular membranes (Peters et al., 1995). In contrast, the GTP-binding defective mutant, ARF6/T27N, localized exclusively to tubulovesicular structures. Since CD treatment redistributes wild-type ARF6 from the PM to tubular structures, we examined whether these structures represent the compartment where ARF6/T27N resides. ARF6/T27N localized to tubular structures originating from the perinuclear region as well as to vesicular structures scattered throughout the cell (Fig. 1). These observations suggest that CD treatment of cells expressing wild-type ARF6 causes a shift in the distribution of ARF6 to the tubular compartment where ARF6/T27N localizes.

The ability of CD to enhance the tubular morphology of the ARF6 compartment was reminiscent of the effects of BFA on increasing the tubular morphology of the transferrin receptor endosomal compartment (Hunziker et al., 1991; Lippincott-Schwartz et al., 1991; Wood et al., 1991). To investigate the relationship between these two endosomal compartments, we examined the distribution of ARF6 and transferrin in HeLa cells transfected with ARF6-HA. The cells were incubated with transferrin alone or in the presence of CD (1 μM), BFA (1 μM), or CD plus BFA for 30 min, and were then rinsed in media, fixed and processed for immunofluorescence with antibodies to HA to detect ARF6-HA-transfected cells and with antibodies to transferrin to detect cell-associated transferrin. In untreated cells, transferrin was localized to the perinuclear region and to punctate structures in the periphery, but showed little overlap with ARF6-labeled structures (Fig. 2, untreated). The distribution of transferrin in transfected cells was indistinguishable from that observed in untransfected cells. In the presence of CD, no change in the distribution of transferrin was observed, whereas the ARF6 distribution shifted towards the tubular structures emanating out of the perinuclear region (Fig. 2, CD). After treatment with BFA, the transferrin compartment exhibited tubular structures, but the ARF6 compartment was unaffected and did not colocalize with the transferrin compartment (Fig. 2, BFA). Treatment of cells with both CD and BFA resulted in tubulation of the transferrin endosome and enhanced tubular morphology of the ARF6 endosome, with little overlap between the two systems (Fig. 2, CD+BFA). The ARF6 compartment also did not colocalize with LAMP, a marker for late endosomes and lysosomes (Mellman, 1996), nor with other organelle markers tested, including those recognizing the Golgi complex and ER (data not shown).

Although ARF6 redistributed from the PM to the tubular membrane compartment during CD treatment, the mechanism of transfer between these two compartments was not clear. Since ARF1 cycles between the cytosol and Golgi membranes (Donaldson and Klausner, 1994), it was possible that ARF6 might also move from the PM via the cytosol to the tubular compartment. However, given that immuno-EM localization (Peters et al., 1995) and subcellular fractionation studies (Cavenagh et al., 1996; Song, J., and J. Donaldson, unpublished observations) suggest that the majority of ARF6 in cells is membrane associated, another possibility was that ARF6 might move between the PM and the tubular compartment by a membrane-mediated process. To begin to address this, we examined whether other plasma membrane proteins might also redistribute to the tubular compartment during CD treatment.

We used Tac, the IL-2 receptor α subunit, as a PM marker. Tac has been thoroughly studied and used as a generic integral PM protein with no retention or targeting information that moves through the secretory pathway to the PM (Leonard et al., 1984; Weissman et al., 1986; Bonifacino et al., 1990; Subtil et al., 1997). We cotransfected HeLa cells with plasmids encoding Tac and wild-type ARF6, and determined the distributions of the proteins after incubation in the presence or absence of CD or AlF. Tac and ARF6 showed extensive colocalization at the plasma membrane and with an internal, juxtanuclear compartment in untreated cells (Fig. 3). Whereas CD treatment of cells shifted the localization of Tac and ARF6 to the tubular compartment, addition of AlF (obtained by adding 30 mM NaF and 50 μM AlCl₃) resulted in colocalization of ARF6 and Tac in surface protrusions (Fig. 3). In HeLa cells coexpressing ARF6 and a Tac chimeric protein containing the cytoplasmic tail of HLA-DM, a sequence specifying clathrin-coated pit localization (Marks et al., 1995), the chimeric protein did not codistribute with ARF6 in either surface protrusions or the tubular compartment (not shown). These results demonstrate that PM proteins lacking known signals for clathrin-coated pit localization colocalize with ARF6 in the tubular compartment of HeLa cells during CD treatment.

The movement of Tac and ARF6 into the tubular endosomal compartment in transfected cells prompted us to examine whether an endogenous PM protein would redistribute to such a tubular compartment with or without ARF6 overexpression. MHC class I proteins (MHC-I) are found on the PM and apparently do not contain cytoplasmic tails conferring clathrin/AP-2 localization (Neefjes et al., 1990). We analyzed the distribution of endogenous MHC-I in HeLa cells, some of which were overexpressing ARF6 by transient transfection, and thus labeled with the antibody to ARF6. In untreated cells, MHC-I localized to the PM, the juxtanuclear region, and occasionally in fine tubular arrays that radiated from this region (Fig. 4, arrowheads). After CD treatment, there was an apparent increase in MHC-I labeling of these tubular membranes that extended out to the peripheral edge of the cells (Fig. 4, CD). The distribution of MHC-I was similar whether the cells were overexpressing ARF6 (detected with the ARF6 antibody) or were untransfected. In the transfected cells, ARF6 colocalized with MHC-I at the PM and in the tubular structures. These observations demonstrate that the tubular compartment is normally present in HeLa cells, becomes more elaborate in the presence of CD, and is not induced by overexpression of ARF6. Thus, we can monitor the movement of ARF6 and Tac in transfected HeLa cells to study this novel membrane system.

Although ARF6 and Tac colocalized in the tubular compartment with CD or at surface protrusions with AlF, it was difficult to discern the ARF6- and Tac-localized structures in the juxtanuclear region. However, it is reasonable to assume that some of this labeling may represent Tac in the Golgi complex. Indeed, a tight juxtanuclear structure labeled with Tac antibody is observed in some cells. This raised the issue of whether the Tac in the tubular compartment was newly synthesized Tac en route to the PM or surface Tac redistributing from the PM.

Therefore, we assessed the movement of surface Tac from the PM into the tubular compartment in cells coexpressing ARF6 and Tac by monitoring Tac antibody internalization. Surface Tac was labeled by incubating cells at 4°C with anti-Tac antibody. The cells were then washed and incubated at 37°C in the presence or absence of CD to allow internalization of the bound antibody. After fixation, the distribution of the antibody-bound Tac was revealed by secondary antibody staining. Localization of the Tac antibody after binding at 4°C showed only staining of the cell surface, where it colocalized with PM-associated ARF6, but ARF6 was also observed in the internal perinuclear and tubular structures (Fig. 5, 4°C Binding). Subsequent incubation of cells at 37°C in the absence of CD revealed that Tac antibody was internalized into the perinuclear and tubular compartment in many cells, where, it then colocalized with ARF6; some of the Tac antibody labeling also colocalized with ARF6 staining at peripheral PM sites (Fig. 5, 37 °C No drug). This observation indicated that Tac normally moved from the PM to the tubular compartment in the absence of any perturbants. Warming the cells in the presence of CD resulted in a more dramatic redistribution of the Tac antibody and ARF6 to the tubular compartment, with little Tac antibody or ARF6 localization at the peripheral edges of the cells (Fig. 5, 37 °C CD).

To assess Tac antibody remaining on the surface after CD treatment, the fixed cells were incubated with the secondary antibodies in the absence of detergent permeabilization. After CD treatment, the Tac antibody remaining at the PM was evenly distributed along the cell surface (Fig. 5, CD, Surface only). No staining of the tubular compartment was observed, indicating that the Tac antibody taken up into the tubular compartment was inaccessible to secondary antibodies and thus not continuous with the PM. The internalization of surface Tac was not induced by antibody cross-linking during the course of the experiment, since the localization of Tac in the tubular compartment was also observed by immunofluorescence after fixation (see Fig. 3). In contrast to the internalization of PM proteins, this compartment could not be loaded with fluid phase markers (not shown). Unlike the movement of surface Tac into the tubular compartment in either the presence or absence of CD, when cells were incubated with AlF at 37°C, the Tac antibody localized to the surface protrusions along with ARF6 and did not redistribute into the tubular compartment (Fig. 5, AlF).

Since surface Tac appears to move from the PM to the tubular compartment, and CD treatment causes Tac to accumulate in this compartment (see above), we wanted to determine whether internalized Tac can recycle from the tubular compartment back out to the PM. To study the fate of internalized Tac antibody, we had to remove the anti-Tac antibody that remained at the cell surface after internalization at 37°C. Rinsing cells with low pH buffers containing high salt has been shown to remove transferrin bound to its receptor at the cell surface (Klausner et al., 1983).

To determine the efficiency of removing surface-bound antibody with low pH buffer, cells expressing ARF6 and Tac were incubated with Tac antibody at 4°C to label surface Tac. Some cells were warmed to 37°C to allow internalization in the presence of CD. The cells were then either fixed immediately (Fig. 6, Total) or rinsed briefly with low pH buffer (0.5% acetic acid, 0.5 M NaCl, pH 3.0; Fig.6, Internal) before fixation and immunolabeling for ARF6 and Tac antibody. The Tac antibody labeled only the cell surface at 4°C, whereas ARF6 localized both to the PM and to internal structures (Fig. 6, 4°C Total). Rinsing with low pH buffer completely removed the surface-bound Tac antibody, but did not alter the distribution of ARF6 (Fig. 6, Internal). Incubation of cells at 37°C for 30 min in the presence of CD resulted in the internalization of Tac antibody into the tubular compartment (Fig. 6, 37 °C +CD Total). Washing with low pH buffer before fixation removed the surface Tac antibody, but did not remove the internalized Tac antibody that colocalized with ARF6 in the tubular compartment (Fig. 6, 37 °C +CD Internal). The inability of the low pH wash to remove Tac antibody lends additional support to the argument that this compartment is internal and not continuous with the PM. The tubular compartment had a beaded appearance possibly because of the low pH wash. Nevertheless, ARF6 and Tac remained colocalized in this compartment.

Having demonstrated that the low pH wash could efficiently remove surface-bound Tac antibody, we examined whether internalized Tac recycled from the tubular compartment back to the PM. Cells were labeled with Tac antibody at 4°C and then warmed to 37°C in the presence of CD to accumulate Tac antibody in the tubular compartment. Antibodies remaining at the cell surface were then removed with a low pH wash at 4°C (as in Fig. 6), leaving only the internalized Tac antibody (Fig. 7, Load).

To determine whether the internalized Tac could recycle back to the PM, the loaded cells were subsequently incubated at 37°C in the presence or absence of CD for 30 min. Tac antibody that reappeared on the surface was detected by labeling fixed cells with fluorescent secondary antibody without detergent permeabilization; ARF6 was subsequently localized in these cells after permeabilization. The internalized Tac antibody reappeared on the surface at protrusive sites along the edges of cells, together with ARF6, in the absence of CD (Fig. 7, −CD). Phalloidin labeling of these cells indicated that the sites of Tac reappearance were enriched in actin filaments (not shown). In contrast, incubation with CD prevented the reappearance of the internalized Tac antibody back on the cell surface (Fig. 7, +CD); ARF6 remained in the tubular compartment and colocalized with the internalized Tac antibody in samples that were detergent permeabilized to visualize total Tac antibody staining (not shown).

These results demonstrate that surface Tac moves into and back out of this tubular endosomal compartment along with ARF6. In the presence of AlF, internalization into this compartment is blocked (Fig. 5), whereas in the presence of CD, recycling of Tac back out to the PM is inhibited.

The distribution of wild-type ARF6 in cells treated with AlF and CD mimics the localization of mutant ARF6 in cells expressing the active ARF6/Q67L and inactive ARF6/T27N mutants, respectively (Radhakrishna et al., 1996; Fig. 1 this study). Thus, we examined whether the blocks in transport observed with these drugs on the movement of Tac between the PM and the tubular endosomal compartment could be recreated in cells expressing the ARF6 mutant proteins.

We first examined the uptake of Tac antibody into cells cotransfected with plasmids encoding Tac and ARF6/ Q67L, following the protocol used in Fig. 6. After the binding at 4°C, Tac antibody was uniformly localized along the PM (Fig. 8, Binding). Upon warming to 37°C in the presence of CD, no change in the distribution of Tac antibody was observed (Fig. 8, CD Total). Washing these cells with low pH buffer before fixation removed all of the Tac antibody (Fig. 8, CD Internal), indicating that the Tac antibody remained at the cell surface during the 37°C incubation. The same results were observed when these cells were warmed in the absence of CD (not shown). This indicates that expression of the active ARF6/Q67L alone blocks the internalization of surface Tac, suggesting that GTP hydrolysis by ARF6 is required for Tac internalization.

We next examined both the internalization and recycling of Tac antibody in cells cotransfected with plasmids encoding Tac and ARF6/T27N. After binding at 4°C, Tac antibody localized to the surface of cells and did not colocalize with ARF6/T27N (Fig. 9, 4°C Binding). Cells were then incubated at 37°C in the absence of CD, followed by low pH removal of remaining surface antibody to assess the internalized Tac antibody. The Tac antibody accumulated in the tubular compartment, where it colocalized in part with ARF6/T27N (Fig. 9, Uptake). After the low pH wash, cells that had taken up the Tac antibody were then further incubated at 37°C to allow recycling, and surface reappearance of Tac antibody was determined as described above (Fig. 7). The internalized Tac antibody did not appear on the PM in these cells (Fig. 7, Surface Reappearance). This indicates that expression of the inactive ARF6/T27N mutant (ARF6-GDP) alone, in the absence of cytochalasin treatment, blocks the recycling of internalized Tac from the tubular compartment back to the PM, suggesting that activation of ARF6 is required for exit from this compartment.

In this study, we describe a novel PM–endosomal recycling pathway that is regulated by the ARF6 GTP-binding protein. We previously showed that the distribution of the wild-type protein could be acutely shifted to the active PM location by treatment of the cells with AlF; dynamic, actin-rich protrusions were induced in these cells as a consequence of this treatment (Radhakrishna et al., 1996). We now demonstrate that inhibitors of actin polymerization, such as CD, shift the distribution of the protein from the PM to an internal, juxtanuclear compartment that exhibits tubular elements extending out towards the periphery. In HeLa cells, this compartment is distinct from the transferrin receptor endosome, and resembles the compartment where the GTP-binding defective mutant of ARF6, T27N, resides. Along with ARF6, “generic” PM, including membrane proteins that do not contain clathrin/AP2 localization domains such as Tac, normally moves from the PM into this compartment and back out again. Although we had difficulty discerning the morphology of the compartment in the juxtanuclear region, it was clear during the internalization experiments that Tac would appear in both the juxtanuclear and tubular portion of this compartment simultaneously, suggesting that they were connected. The two treatments that alter ARF6 distribution also result in specific blocks in this membrane traffic pathway. CD inhibits egress of ARF6 and membrane out of this system, whereas AlF prevents the internalization of ARF6 and membrane into this system. These transport blocks observed pharmacologically with CD and AlF treatment were recreated by expression of either the inactive T27N or the active Q67L mutants, respectively. This demonstrates that the membrane recycling pathway described here is regulated by the GTP cycle of ARF6.

A recent study reported that endogenous ARF6 is associated with a PM fraction isolated from CHO cells and is not associated with early endosomal membranes, identified by loading with the fluid phase marker HRP (Cavenagh et al., 1996). In transfected HeLa cells, we observe that ARF6 is associated with an internal membrane compartment in addition to the PM. This compartment is accessible to loading with surface PM proteins, is transferrin negative, and does not accumulate fluid phase endocytic tracers. Accordingly, this ARF6-associated membrane compartment that we observe would not be detected as an “early endosome,” but rather, would likely be included in the PM fraction isolated by Cavenagh et al. (1996).

We propose a tentative working model for how ARF6 regulates this membrane recycling pathway (Fig. 10). Through its GTP cycle, ARF6 moves between the PM and this recycling compartment. Nucleotide exchange onto ARF6-GDP is required for ARF6 and membrane to exit from the recycling compartment. The reappearance of ARF6-GTP and membrane at peripheral exit sites correlates with the formation of actin-based protrusions. Hydrolysis of the GTP bound to ARF6 is required to allow internalization of PM and ARF6-GDP back into this internal compartment. The specific mechanism whereby the membrane is internalized, i.e., whether by nonclathrin pinocytosis or tubular membrane invaginations that become discontinuous with the PM, is yet to be identified. The internal endosomal compartment includes ill-defined structures in the juxtanuclear region and associated tubular elements that extend out to the periphery. We speculate that treatment of cells with CD blocks ARF6–GTP exchange and membrane recycling, and that treatment with AlF blocks ARF6-GTP hydrolysis and membrane internalization. We are currently developing methods to investigate the nucleotides bound to ARF6 under these different conditions, and to further characterize the membrane intermediates involved in the movement into and out of this recycling compartment.

In some respects, the requirement in our model for ARF6 activation (i.e., GTP exchange) for membrane recycling to occur is similar to the observations made by D'Souza-Schorey et al. (1995) for the transferrin cycle in CHO cells. In their study, overexpression of either the wild-type ARF6 or the constitutively active Q67L mutant inhibited transferrin uptake into CHO cells, whereas expression of the T27N mutant inhibited transferrin receptor recycling back to the cell surface. Several distinctions between the two studies, however, deserve comment. First, the extent to which this endosomal recycling pathway regulated by ARF6 is separate and distinguishable from the transferrin receptor endosome, as we observed, may vary among different cell types. Indeed, in CHO cells, the transferrin receptor and the T27N mutant of ARF6 colocalize (D'Souza-Schorey et al., 1995), whereas in HeLa cells, the compartment where ARF6 localizes is clearly distinct in its morphology and behavior from the transferrin receptor compartment (Radhakrishna et al., 1996; this study). Second, we were able to confirm the existence and behavior of the ARF6-regulated compartment in HeLa cells in the absence of ARF6 overexpression by observing endogenous MHC-I in untransfected cells. Finally, in CHO cells, overexpression of the wild-type ARF6 alone causes a block in the internalization of transferrin receptors (D'Souza- Schorey et al., 1995). The inability to reverse this effect of the wild-type protein limits the use of this system for defining the role of the ARF6–GTP cycle in this process. By contrast, in HeLa cells we were able to study the full PM– endosomal recycling pathway regulated by ARF6, since overexpression of the wild-type protein, although exhibiting enhanced nascent protrusive structures at the cell surface, does not inhibit membrane movement through the pathway. This has allowed us to identify pharmacologic treatments that acutely and reversibly alter the distribution and movement of ARF6 and membrane proteins through this pathway.

In HeLa cells, the ARF6-regulated recycling compartment has a distinctive tubular morphology that emanates out of the juxtanuclear area near the microtubule organizing center. Tooze and Hollinshead (1991) were among the first to document the existence of extensive tubular endosomal networks observable by thick-section EM in a variety of cell types, including HeLa cells. Although in some cases tubular endosomes have been reported to contain the transferrin receptor (van der Sluijs et al., 1992; Hopkins et al., 1994), the ARF6-regulated tubular endosome that we observe does not. There are other reports of novel endosomal compartments distinct from the well-characterized transferrin endosomal system. The autocrine motility factor receptor localizes to a tubular compartment in MDCK and HeLa cells that does not colocalize with transferrin receptor (Benlimame et al., 1995). In NRK cells, expression of endotubin, a protein normally associated with the apical early endosomes of intestinal epithelia, localizes to a compartment that contains neither transferrin nor fluid phase markers and is not altered by treatment with BFA (Wilson and Colton, 1997). It will be interesting to determine whether either of these proteins are associated with the ARF6-regulated membrane compartment described here.

Although the tubular morphology is a hallmark of the ARF6-regulated membrane compartment in HeLa cells, this morphology may not be exhibited by ARF6-regulated compartments in other cells. For example, it was recently reported that ARF6 cofractionates with isolated chromaffin granules and may be involved in the regulated exocytosis of these granules (Galas et al., 1997). Although these granules do not exhibit tubular morphology, through its GTP cycle, ARF6 may still regulate membrane movement through such compartments.

Having demonstrated the existence of a membrane recycling pathway regulated by ARF6, one might ask what function such a pathway serves in the cell. We showed that bulk PM was capable of being internalized and recycled out of such a compartment, and that this apparently occurs on a slower time scale (loading and unloading of this compartment on the order of ∼30–60 min) than the transferrin receptor internalization and recycling pathway (loading and unloading on the order of 5–10 min; Mellman, 1996). The ARF6-regulated pathway might allow for bulk membrane to be internalized and then recycled to the cell surface at defined sites in such a way that it could be used to regulate the surface area and shape of cells. Indeed, in HeLa cells, ARF6 and membrane recycle to defined sites along the PM often associated with actin-rich protrusions. Additionally, we have demonstrated that recycling of ARF6 and membrane to the edge of cells is required during cell spreading; expression of ARF6/T27N inhibits HeLa cell spreading (Song, J., Khachikian, H. Radhakrishna, and J. Donaldson, manuscript in preparation). These observations suggest that the ARF6-regulated membrane cycle may be involved in PM remodeling events that are initiated in response to physiological stimuli (e.g., growth factors, chemotactic agents, and metastasis).

It is intriguing that the trafficking and possibly GTP status of ARF6 are influenced by actin polymerization. The observation that recycling of ARF6 and membrane back to the PM is blocked by CD treatment or expression of the T27N mutant suggests that membrane movement out of the tubular compartment requires actin polymerization and ARF6-GTP. At the present time, we cannot determine whether the requirement for actin polymerization is needed for nucleotide exchange, and thereby conversion to ARF6-GTP, or if it is required for ARF6-GTP and membrane to recycle back to the PM, for example, through an actin-myosin based transport step (Titus, 1997). It is not likely that CD is specifically inhibiting nucleotide exchange onto ARF6, analogous to the effects of BFA on ARF1 at the Golgi (Donaldson et al., 1992b; Helms and Rothman, 1992; Randazzo et al., 1993), since other inhibitors of actin polymerization, including cytochalasin B and latrunculin, also inhibit movement of ARF6 back to the PM. Further details of the actin dependence of this step will require in vitro methods to analyze the nucleotide status of ARF6 and reconstitute the ARF6-mediated membrane recycling to the PM. Once at the PM, the accumulation of ARF6 and perhaps the delivery of membrane there results in a stimulation of actin polymerization and formation of protrusive structures. Thus, it appears that ARF6 (localization/GTP status) depends on, but is also capable of stimulating, actin polymerization. Further studies should elucidate the connection between this novel ARF6-regulated membrane traffic pathway and the actin cytoskeleton.

We thank Drs. J. Bonifacino, M. Marks, and P. Roche for generously providing reagents used in this study. We also thank Drs. O. Al-Awar, E. Korn, J. Lippincott-Schwartz, and A. Sobota for critical reading of the manuscript.

ARF

ADP-ribosylation factor

AlF

aluminum fluoride

BFA

brefeldin A

CD

cytochalasin D

HA

hemagglutinin

MHC-I

MHC class I proteins

PM

plasma membrane